OpenAI has introduced ChatGPT Agent, a powerful new tool designed to perform complex, multi-step tasks on a user’s behalf by operating a virtual computer. This marks a major step forward from traditional chatbots, giving users an AI-powered assistant capable of managing tasks like scheduling, researching, shopping, and even creating presentations.
During a demonstration, OpenAI’s product and research leads, Yash Kumar and Isa Fulford, explained that ChatGPT Agent uses a new model trained with reinforcement learning on tasks that involve multiple tools — including a text browser, visual interface, and terminal. It combines features from OpenAI’s existing tools, Operator and Deep Research, and its development brought together teams from both projects.
Examples shown during the demo included planning a dinner date by syncing with Google Calendar and OpenTable, creating a research report comparing toys, and automating personal tasks like weekly parking requests at OpenAI’s office. Fulford shared that the tool performs better than previous systems when used for online shopping, thanks to the combined capabilities of its foundational tools.
Although the tool can be slow at times, the team emphasized that its strength lies in completing tasks that would otherwise take much longer if done manually. Users can run tasks in the background and return later to check the results. Importantly, the agent always requests permission before taking irreversible actions like sending emails or making bookings.
To ensure safety, OpenAI has implemented security measures around biological and chemical capabilities, even without evidence of potential misuse. Financial transactions are currently restricted, and a feature called Watch Mode ensures the tool ceases activity if users navigate away from specific tabs during sensitive operations.
Starting today, ChatGPT Agent is rolling out to Pro, Plus, and Team users, with availability for Enterprise and Education users expected later this summer. However, no launch date has been set for users in the European Economic Area and Switzerland.
This launch is part of a broader industry push toward advanced AI agents, with OpenAI joining companies like Google, Amazon, and Anthropic in racing to develop digital assistants that function more like virtual coworkers than simple chatbots.
